on the role of microeconomics, planning, and finances in ... the role... · on the role of...

PROCESS SYSTEMS ENGINEERING

On the Role of Microeconomics,Planning, and Finances in Product Design

Miguel J. BagajewiczChemical, Biological and Materials Engineering Dept., University of Oklahoma, Norman T-335, OK 73019

DOI 10.1002/aic.11332Published online October 16, 2007 in Wiley InterScience (www.interscience.wiley.com).

In this article it is claimed that in order to obtain the right composition, structure,and functionality for a new product, one needs to anticipate what the demand of theproduct will be and take into account all costs involved (associated manufacturing andsupply chain). To obtain the demand, one needs a pricing model, which in turn relieson consumer preferences that are connected to the product composition, structure, andfunctionality. Thus, a model, including the varying characteristics of the product, themanufacturing capacity and site, the supply chain and ultimately, the markets, isproposed. I use a very simple case of insect repellent to illustrate how the best repel-lent, identified as the one that consumers will prefer the most, is not the most profita-ble one and how one can obtain the insect repellent composition that maximizes profit.� 2007 American Institute of Chemical Engineers AIChE J, 53: 3155–3170, 2007

Keywords: optimization, product design

Introduction

Product design has been advocated to be one of the newfrontiers opened for chemical engineers.1,2 It is claimed thatwe are moving from a commodity based chemical industry to ahigh value added and product performance-based one. Somecall it a shift in interest,3 with obvious impact in research andeducation,2,4 while others advocate that this is just an expan-sion of the competency that will still include the commoditysupply chain, but will incorporate the new performance-basedconstraints that a product contributes.5–10 Bagajewicz11,12

claims that this expansion goes farther than defining productperformance and that the definition of a venture associated to aparticular product goes all the way from its molecular design,which settles its properties, to all the finance aspects (commer-cialization, pricing, etc), going through the definition of themanufacturing process and the associated supply chain.

This paradigm ‘‘shift’’ or paradigm ‘‘expansion,’’ suggests,arguably implicitly, that process engineering is a maturefield, while product design is a relatively virgin filed, at leastvirgin from the use of tools and methods that the PSE com-

munity. One good example of efforts following the suggestedpath is the article by Wibowo and Ng,13 who analyze theissues associated with the fabrication of creams and pastes.There are several papers dealing with refrigerant develop-ment,14,15 drug development,16–18 solvent, and computeraided molecular design in general,10,19 all of which are infact product design after all. Cussler6 illustrates some of thedifferences between the two paradigms (Table 1). To do thecomparison he uses the so-called Conceptual Design para-digm developed by Douglas,20 which is very similar tothe Onion Model proposed by Smith and Linnhoff.21,22 Thereare, of course, other approaches to process design, likethe reducible superstructure approach (best represented bythe book of Biegler et al.23) for which we present our owncounterpart version in this article.

The table has a couple of interesting features

(a) Makes product design the center, albeit in the sense of‘‘discovery’’ of existing commercial molecules, but it alsoapplies to molecular design.

(b) Suggests ad hoc idea generation and selection stepsthat presumably vary from product to product, for which asystematic search is not available or has to be constructedcase by case.2 We have seen examples on searches driven by

Correspondence concerning this article should be addressed to M. J. Bagajewiczat [email protected].

� 2007 American Institute of Chemical Engineers

AIChE Journal December 2007 Vol. 53, No. 12 3155

functionalities (refrigerants, drugs, etc.) like those performedby Camarda and Maranas,24 Sahinidis and Tawarlamani,15

and also described in Ostrovsky et al.25 and Achenie andSinha.19 Recently, Hill3 suggested that the issue is not compo-sition alone anymore but it also involves microstructure.

(c) Some who advocate product design2 only call it‘‘chemical’’ product design, which rules out mechanical andelectronic and electromechanical devices, etc. Therefore, it isonly a matter of time until this expands to all products.Evans,26 for example, has recently emphasized the upcomingintegration between the process industries as providers ofcommodities and the discrete industries the providers ofpackage goods, devices, appliances, automobiles, etc.

(d) The title of the table suggests that these are oppositeand to an extent, excluding activities.

Gani27 has reviewed the challenges of the field point out the‘‘needs for multilevel modeling with emphasis on propertymodels that are suitable for computer-aided applications, flexi-ble solution strategies that are able to solve a large range ofchemical product design problems and finally, a systems chemi-cal product design framework with the overall objective toreduce the time and cost to market a new or improved product.’’

Stephanopoulos5 reemphasized the idea of product designand suggested that manufacturing is indeed migrating frombeing process (commodity)-centric to being product-centric,all this judged by the performance of the companies in thestock market and other associated facts. He suggests that acompany should maximize value-added through the supplychain and that while the process centered industry focuses oncommodities, the product-centered industry focuses on identi-fication of customer needs as a driving force. He asks if‘‘Process Systems Engineering is prepared to engineer(design and manufacture) products, or someone else shoulddo it?’’ The question was provocative enough to rise my in-terest and help showing that this IS very much a problem tobe handled with PSE tools, although, as I describe in this ar-ticle, extending the hands to get help from disciplines thathave not been traditionally its partners. In addition, one mustwonder if he refers to the role of chemical engineers in allend user products, not only chemical products. This is sup-ported by calls to extend the role of the PSE community intoformulations.28

In parallel to this push for extending the borders of chemi-cal process systems engineering to new areas, a new conceptof supply chemical chain was recently discussed in detail byGrossmann and Westerberg29 and later by Grossmann30 andin the United States National Academies report by Breslowet al.31 The chemical supply chain extends from the moleculelevel to the whole enterprise. Breslow et al.31 suggest that(bold and underlining was added).

‘‘Another important aspect in the modeling and optimiza-tion of the chemical supply chain is the description of the

dynamics of the information and material flow through thechain. This will require a better understanding of the integra-tion of R&D, process design, process operation, and busi-

ness logistics. The challenge will be to develop quantitativemodels that can be used to better coordinate and optimizethe chemical enterprise. Progress will be facilitated by newadvances in information technology, particularly throughadvances in the Internet and by new methods and mathemati-cal concepts. Advances in computer technology will play acentral role. Fulfilling the goal of effectively integrating thevarious functions (R&D, design, production, logistics) in thechemical supply chain will help to better meet customer

demands, and effectively adapt in the new world of e-com-merce. Concepts related to planning, scheduling, and controlthat have not been widely adopted by chemical engineersshould play a prominent role in the modeling part of thisproblem. Concepts and tools of computer science and opera-tions research will play an even greater role in terms ofimpacting the implementation of solutions for this problem.’’

Hill3 presented an overall product design procedure startssimilar to that of Cussler.6 He identifies the following steps

(a) Identification of consumer ‘‘needs.’’(b) Translation of the need in a technical target (quantifi-

cation of the need).(c) Reduction to a physical prototype, that is, the identifi-

cation of what substance or set of substances (active ingre-dients) that could help accomplish the target property.

(d) Testing a physical prototype. Actually manufacture thenew product and test it against a variety of relevant criteria.

Hill3 claimed that chemical engineers are involved withsteps (c) and (d) but hardly with steps (a) and (b). In addi-tion, he claims that to the process above described, one needsto add the assessment of the manufacturing process (thiswould be step (e)).

Hill3 also argued that mathematical programmingapproaches, which exist for certain combinations of theabove process are not yet available for the case of a struc-tured product, where aside from composition, other physicalconsiderations (texture, microstructure, stability, etc) are notcovered.

As it has been pointed out earlier, models that target mo-lecular discovery through mathematical programming basedon targeted properties exist. Moreover, attempts to performmolecular discovery through mathematical programming to-gether with process synthesis have been developed.

It is of course well known that steps (a) and (b) involvemarketing considerations such as anticipated demand, tenta-tive prices, consumer behavior, advertisement costs, possiblycontracts, etc. While Hill3 hints that these two steps (a andb) are the steps that generate targets for the rest of the steps,he does not elaborate on all the marketing issues, implying, Ipresume, that they are taken care somehow on their own. Inaddition, connections between manufacturing and marketsthrough supply chains like plant location and pricing has alsobeen developed.32

Regarding marketing, the term coined to describe the pro-cess of bringing a new product service to market is newproduct development (NPD). There is even a Product devel-opment and management association (http://www.pdma.org/about/) that nucleates practitioners around workshops, publi-cations, and conferences. All soft issues concerning new

Table 1. Process Design vs. Product Design(Following Cussler (Ref. 6))

Process Design Product Design

Batch versus continuous Customer needInput/output Idea generationRecycles SelectionSeparation/heat Manufacture

3156 DOI 10.1002/aic Published on behalf of the AIChE December 2007 Vol. 53, No. 12 AIChE Journal

products are covered in the handbook edited by Kahn.33 TheNPD process is conceived as composed of two traits: ideageneration, product design, and detail engineering on oneside and market research and analysis on the other. Theseideas of integration over the whole enterprise linking busi-ness decision-making to process and product design havebeen also substantiated by Ng.34

In this article I propose a modeling scheme for productdesign, which counterparts what the reducible superstructureapproach is for process design. Moreover, I claim that thereis no such thing as the differentiation established in Table 1.Rather, I claim, process design is one (important) part ofproduct design. In addition, as it will become apparent in thearticle, I claim that without considering microeconomics,product design cannot be properly accomplished. The ques-tion remains at what point in the cycle, microeconomicsneeds to be considered. I claim that in some form, it shouldbe included immediately. In the rest of the article, I discussthe different types of products, differentiating between com-position, structure, and functionality. Then, the current stateof the art is described followed by the proposed model. Afterthe model is presented, I discuss the role of pricing theory,and later review the methodologies already available for inte-gration with finances, supply chain design/operations. A sim-ple example is presented to illustrate some of the novelideas.

Consumer-Related Properties of Products

One group of products are pure components or mixtures,in various forms (sometimes structured), like creams, pastes,lotions, soaps, food, flame retardants, insect repellents, pesti-cides, etc, which are either directly consumed by end-usersor sold as raw materials for other end-user products. Theircharacteristic is that they are ‘‘used’’ once. The other groupof products is devices, which serve certain functionality andare characterized by the fact that can be used repeatedly.They are in general, mechanical or electronic devices.

Although the boundary between these two groups is notwell defined sometimes, both groups have been recognizedby Chemical Engineers as excellent opportunities in whichthey could make a difference, even in cases where there aremany other engineering disciplines involved. One example isthe hemodialysis device, the fuel cell, or the espresso coffeemaker.4

One needs to be careful in defining ‘‘targets’’ for thedesign of these products, especially if language that will beshared by different fields (marketing, economists, and engi-neering) is to be used. Thus, in defining a product, one can-not at first express its properties and functionalities using en-gineering-like properties like viscosities, heat capacities,vapor pressures, etc, as well as device functionalities likeefficiencies, etc. One needs to identify other ‘‘properties,’’those that make customers ‘‘happy’’ or ‘‘satisfied.’’ Thus, forexample, soaps and lotions have properties like ‘‘creaminess,thickness, and smoothness,’’2,13 which are related to viscosityand surface tension and if the lotion/soaps/cream is multi-phase to the proportion of phases, drop sizes, etc. Foodshave all types of ‘‘tastes’’ that can be related to the concen-tration of substances (sugar, pepper, species, additives ingeneral), and ‘‘textures.’’

I call these ‘‘consumer properties’’ (y), with the under-standing that they need to relate through some continuousfunction g(l) to the engineering-properties (viscosities, den-sities, composition, membrane area/thickness, etc), denotedby x.

State of the Art and Challenges

There are two types of product design activity� Product design through combinatorial enumeration and

optimization of existing (a-priori defined) alternatives.� Product ‘‘discovery’’ and design through mixture (or

even molecular) design. We now need to add structuredesign to achieve functionality.

The former makes use of simple enumeration of existingcandidates, using certain sorting criteria. One would call thissome sort of ‘‘selection’’ process, in contrast with the discov-ery associated to the design of novel products which havenever been synthesized. In the case of products composed ofmixtures of components like soaps, margarine, etc., as citedin Hill3 the associated issue is composition and structure.

Discovery is more difficult because it requires that aproperty prediction scheme be built associated with a set of‘‘groups’’ or atoms and their interconnection pattern thatwill constitute a molecule or a mixture of molecules. In thiscase, the synthesis step requires the identification of exist-ing commodities and their proportion in a mixture, and/or aset of reactions that would lead to new molecule(s). Exam-ples of this type of procedure can be found in severalarticles and books.4,35 For example, the ‘‘discovery’’ of newrefrigerants can be made targeting high heat of vaporizationand low heat capacity; Cussler and Moggridge2 do the samefor de-icing products. In the case of devices, Seider et al.4

discuss the hemodialysis machine or the automotive fuelcell. When ‘‘market driven’’ new products are considered,the target performance is usually set by Marketing. ‘‘Tech-nology driven’’ new products, where properties are first setby discovery on the engineering side. As we shall see later,in both cases, no matter who identifies the consumer‘‘needs,’’ a consumer model to establish the relationshipbetween demand and price taking into account consumerpreferences is needed. Finally, Costa et al.8 discuss severaladditional intricacies, including the need of modeling con-sumer preferences.

Thus, what marketing provides is one and only one instanceof consumer properties that corresponds to one state of con-sumer ‘‘satisfaction’’ and it also figures out (simultaneously orlater) what price the consumer would be willing to pay for it.In addition, it targets markets in which the product will besold. This identification is based on perceived consumer needsand is obtained using various instruments like surveys as wellas direct exposure of prototypes to consumers. This type ofdata gathering has been extensively studied.36

When engineers analyze the desired properties, assumingthat they are feasible to achieve, they provide the costs, butthey do so by targeting a set of well-known physico-chemicalproperties (x). Somewhere in this process there is a functionx 5 g(y), that translates consumer properties into desiredphysicochemical properties. Once the vector of desired prop-erties x 5 g(y), is known, engineering and R&D provide inreturn the product structure z and the cost through a function

AIChE Journal December 2007 Vol. 53, No. 12 Published on behalf of the AIChE DOI 10.1002/aic 3157

C(l), as illustrated in Figure 1. Associated to the productstructure, Engineering assesses manufacturing and distribu-tion costs, and even might produce prototypes and test them.

Marketing makes use of several different tools to performthe choices of markets and to determine product composi-tion/structure: product positioning, value pricing, etc. Productpositioning37–39 requires defining the market, identifying thedimensions (attributes) of the product, sample customers,contrast with the market preferred combination of attributes(ideal product) and determine the ‘‘position’’, i.e., the dis-tance from ideal. The process also includes creating an imageor identity of the product in the minds of their target market.In turn, value pricing40 is the practice of setting prices basedon the value of a product to the customer. It is customary tomake graphs of perceived price versus perceived benefit.Although perceived benefit is easy to comprehend, perceptionof price, instead of straightforward price is used because usu-ally consumers do not compare prices based on the samequantity/volume/mass of product and there are other factorslike product presentation, packaging, guarantees, etc. The dif-ference between the latter and the former is the value. Thisis illustrated in Figure 2.

A product that is in the value equivalent line (45 degree line)has the ‘‘right price’’ from the point of view of the consumer.This technique can eventually capture the relationship betweenproduct composition/structure (if one is designing a product)and perceived benefits, as well as determining the perceivedprice. Unfortunately, the technique does not allow predictingdemand so some method to determine it needs to be used.

The scheme presented in Figure 1 suggests an iterativecycle where marketing adjusts its desired consumer proper-ties y until the profitability is maximized. I do not know howmany iterations are performed in industry (if any), but evenif they are performed to convergence, one has to wonderhow these iterations are performed. In other words, given ziand therefore Cost (zi), how is yi11 obtained? While this ishard to know and may be it is an ad-hoc procedure in eachcase anyway, what we know is that what Marketing usuallywants is to maximize profit, so that is the starting point tobuild a strategy.

Finally, there are indications that, even when well definedand well-posed, this type of iterative procedure may not beoptimal. Indeed, Guillen et al.41 have looked at a scheme inwhich production scheduling provides costs of multiple prod-

ucts based on demands provided by marketing, which in turnhas its own model to perform pricing, and therefore is able tocome up with the corresponding demand (the simple price 3demand 5 constant relation was used). The idea is based onthe fact that altering the production schedule in multiproductfacilities should (and indeed does) have an effect on the fixedcosts per unit used in existing classical pricing models.42–45

When this model was run iteratively, it was found it some-times diverges and even when converged, it rendered an infe-rior solution to the case in which both models are integratedinto a single one. We might expect the same behavior forproduct design, although I am not willing to try because it isactually likely to be easier to deal with the integrated modelanyway. What this model also showed is that prices should beviewed as first stage decision variables instead of exogenousparameters (unless the firm is a price-taker).

Regarding integration, Breslow et al.31 suggested the needof ‘‘integration of several parts of the chemical supplychain,’’ which ‘‘will give rise to a number of challenges,such as modeling for molecular dynamics, integration ofplanning, scheduling and control (including internet based),and integration of measurements, control, and informationsystems,’’ but fall short of discussing the full integration witheconomics management and business.

Therefore, aiming at the said integration we want to havean integrated model that will help making the followingdecisions simultaneously

� Product structure/composition/functionality� Market choice� Product price for each market� Associated manufacturing method� Associated supply chain, including plant locations and

transportationAssociated to these decisions, one would get:

� Capital investment and manufacturing costs� Supply chain costs� Projected sales (demand) in each market� Revenues, etc.Thus, the newly proposed paradigm for product design

suggests a structure like the one given in Figure 3.Some elements addressing how one could address all these

decision making process is described by Pekny46 whoexplores the role of different algorithm architectures in large

Figure 1. Current state of the art in product design.

[Color figure can be viewed in the online issue, which isavailable at www.interscience.wiley.com.]

Figure 2. Value pricing scheme.


scale engineering problems. In addition, one important con-clusion should be made:

Process Design, which is mistakenly associated to thedesign of production facilities of commodities only, isan integral part of product design. Product design cannotexist without process design. So, there is no antagonismof any sort and distinctions as those of Table 1, shouldonly be made when the limits of the design exercise arerestricted.The task, then, is to construct an integrated model that

will contain several building blocks: A supply chain model,a process synthesis model (or process optimization if theprocess exists and it will be adapted, or process retrofitmodel if changes to existing plants are to be considered), aplanning and scheduling tool if the new product will be partof a family of products manufactured in multiple sites, aconsumer satisfaction and pricing model, and some algo-rithm that will connect product structure/properties/func-tionalities to the other models. Of all these, the first threeare fairly well known. Of the last two, several conceptsfrom microeconomics can be borrowed and used for theconsumer pricing model but the algorithm making theaforementioned connections needs to be built almost ad-hocfor every case. I am nonetheless providing some frameworkon how it could be attempted in conjunction with the pric-ing model.

Model Mathematical Structure

The integrated two stage stochastic model proposed is asfollows. Let w1 be a variable vector containing first stagevariables, such as manufacturing capacity, plant locations,warehouses, transportation means, customer zones, etc. Letw2 be the set of second stage vector of variables, such asproduction levels, demands, advertising expenses, etc., andlet p be the price (assumed a first stage variable for simplic-ity). Finally, using the scenario approach, let ps be the proba-bility of scenario s; let q the scenario independent parame-ters, which are costs of plant equipment and plant construc-tion, parameters needed to calculate the product properties(assumed certain for the time being), etc.; finally, let ts bethe scenario dependent parameters of the model, like futuredemand constraints, the consumer budget (Y), the weights inthe consumer model, etc, the prices of the raw materials and

the utilities, etc. Then we want to maximize the ExpectedProfit as shown below

Maxz;p

Xs

psNPVRs � Fixed Capital Investment

s.t.

NPVRs ¼ Saless �Manufacturing Costss

� Supply Chain Costss �Marketing Costss

Saless ¼ Sales ðz; p;w1;w2s; q; tsÞFixed Capital Investment ¼ Fixed Capital Investment

ðz;w1; ; q)

Manufacturing Costss ¼ Manufacturing Costs

ðz;w1;w2s; q; tsÞSupply Chain Costss ¼ Supply Chain Costs

ðz;w1;w2s; q; tsÞMarketing Costss ¼ Marketing Costs ðz;w1;w2s; q; tsÞ

9>>>>>>>>>>>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>>>>>>>>>>>;

(1)

We called it two stage model because it has ‘‘here andnow’’ decisions (first stage variables) and ‘‘wait and see’’ orrecourse decisions (second stage variables). The former aredecided upfront and the latter are taken in response to certainscenario materializing as illustrated by Barbaro and Bagaje-wicz.47

We have thus reduced the problem of product design todetermining the functional relation of all these function withthe product composition/structure/functionality z. Indeed, themodel should ‘‘discover’’ or ‘‘design’’ a new molecule/mix-ture/structure by adequately varying z, which contains struc-tural molecular parameters (groups, or other information likeconnectivity indices), concentrations of compounds in mix-tures, phases, structure, etc., or it optimizes continuous pa-rameters of a known mixture. Everything, first stage and sec-ond stage costs, as well as second stage sales are thusdependant on the molecular, microscopic design, or the struc-ture/functionality chosen. This is the smallest scale. Thenthere is the manufacturing scale, where the appropriate tech-nology to produce the chemicals/devices is selected, theinvestment level is determined, etc. In the next scale, onehas the supply chain costs (plant location, transportationissues, etc.). Finally, the Marketing scale includes sales andmarketing costs, including advertising.

Process synthesis (flowsheeting) is mature enough to pro-vide means to obtain a flowsheet, and therefore the corre-sponding fixed capital investment as well as manufacturingcosts associated to any product described by z for any sce-nario. The same can be said for methodologies to design sup-ply chains: they are well-known and directly usable.Although I make the assumption here that superstructure-likemethods should be able to capture the manufacturing struc-ture, I recognize that in some cases they may be too hard toformulate. This might be the case for the manufacturing ofcomplex composite materials, thin films, etc. or when molec-ular discovery is part of the design, which requires the reac-tion synthesis path to be added, even in the case of method-ologies used in specialty chemicals, agrochemicals, food,pharmaceutical, bio, and related industries. In particular,pharmaceutical and biomedical devices are intertwined withcomplex networks of HMO’s, insurance companies, etc., that

Figure 3. Proposed product design model.



makes the case rather cumbersome to model. I intentionallyleave this angle of the problem unexplored in this article so Ican advance the concepts surrounding the modeling of theconsumer behavior, which I feel are currently missing alto-gether from the PSE product design literature.

Advertising, and costs associated to sales directly are stand-ard, albeit not simple sometimes. Finally, the sales functionSaless is simply given by the product of price p and demand ds.The demand vector for scenario s, can be determined usingpricing models. This is then, the object of the following sec-tion, where we establish the role f z in the relationship price-demand, which we borrow from microeconomics theory.

Having stated what the concept of the model is, I nowpresent one version of the consumer pricing model and theconsumer preference parameter. Later, I concentrate on illus-trating the calculation of the consumer preference parameterand how the consumer pricing model alters the design of theproduct. The example was chosen in such a way that manu-facturing is not an issue (it is just mixing) and supply chainis oversimplified. I am certain that many other products willexhibit strong manufacturing and supply chain contributions,which I leave for future articles to illustrate. At the end ofthe article I come back to discuss additional issues, likeadvertising, uncertainty, financial risk, budgeting, contracts,etc., to simply indicate how they fit in the above presentedscheme.

Pricing Model

We assume first that there is an established market for thenew product and that what we are looking is for (profitable)substitutes. The question is what price will be the right oneto attract the optimal number of customers and the newdemand associated to it, for a given new product.

We begin with posing the consumer optimization problem.In classical microeconomics, this is posed as follows43,45:Consider two products, with demands d1 (for the new prod-uct) and d2 (for the existing products). Then the consumermaximizes his utility (satisfaction) u(d1, d2) subject to abudget limitation, that is:

Maxd1;d2

uðd1; d2Þs:t:

p1d1 þ p2d2 � Y

9>>=>>;

(2)

where p1 is the new product’s selling price, and p2 the com-petitor’s product price. A typical utility function is concaveand has constant elasticity of substitution, which is a termused to describe the shift from one product to another underprice shifts and is defined as follows:

rES ¼ �@ d1

d2

� �p1p2

� �

@ p1p2

� �d1d2

� � ¼ �@ ln d1

d2

� �

@ ln p1p2

� � (3)

To understand this, consider a certain level of consumptionof the two products d1 and d2 and consider a relative increase

in p1p2

� �of k%. Then, the assumption (rooted in observations)

is that the relative change in the ratio d1d2

� �is krES%. A few

utility functions that satisfy this property have been pro-posed. One well-known function is the Cobb-Douglas utility

u(d1, d2) 5 da1 db2, which has rES 5 21. Because utility

functions are concave, that is, marginal utility decreases withconsumption, a and b are smaller than one. When a 1 b 51the utility doubles when both d1 and d2 double. Conversely,when a 1 b \ 1, utility increases by a factor smaller thantwo when both d1 and d2 double, which means there is socalled decreasing returns to scale, which is standard. TheCobb-Douglas utility has some problems. For example, it iszero when only one product is consumed, which is not true.In addition, the solution to consumer utility maximization is:

p1d1 ¼ abp2d2 (4)

or alternatively

p1d1 ¼ aYðaþ bÞ (5)

which says that revenues p1d1 are constant, no matter what onedoes with prices. If production costs are proportional to d1,then maximization of profit (understood in its simplest form asrevenues minus costs) will suggest d1 5 0 and p1 ?1.

There are many alternatives to the Cobbs-Douglas utility.One that is additive is the Constant Elasticity of Substitution(CES) utility: u(d1, d2) 5 (dq1 1 dq2)

1/q, for which rES 5 1/(q2 1). Note that this function is linear in di when the demandfor the other good is zero. The function is concave for qsmaller than one, which provides diminishing marginal utilityas the demand increases. Finally, the solution to consumerutility maximization is:

p1d1�q1 ¼ p2d

1�q2 (6)

or alternatively

p1d1 ¼ p2ðY � p1d1Þ1�qdq1 (7)

which says that revenues p1d1 are zero for d1 5 Y/p1, andgoes through a maximum. Profit maximization occurs whenp2(Y 2 p1d1)

12q dq1 2 Kd1 is maximum at some point betweend1 5 0 and d1 5 Y/p1 (K is the production cost per unit).Other utilities can be found in standard books of microeco-nomics and pricing theory and using them does not alter theproposed model, as long as they are not inconsistent, like thecase of the Cobb-Douglas utility above illustrated.

So far, we discussed these utilities without mentioning thedifference in quality between products. In fact, the emphasishas been put in the reaction of consumers to prices. In thecase of the Cobbs-Douglas utility, one can think of a and bas related to the respective products’ quality. Indeed, givenequal prices (p1 5 p2), we get d1 ¼ a

b d2 and therefore ab can

be understood as the consumers making choices according topreferences not related to prices. But we have already seenall the problems this utility presents. In the case of CES util-ity, we have the phenomenon that when prices are equal,demands are equal irrespective of the value of q.

To overcome the problems presented by the CES utility,we construct the utility based on functions of demand, which


we could call ‘‘satisfaction functions,’’ that is, we could writeu(d1, d2) 5 (hq1 1 hq2)

1/q, where hi 5 hi(di). Thus, elasticityof substitution is constants with respect to x1 and x2, whichmakes sense. We propose the following:

h1 ¼ ad1 (8)

h2 ¼ bd2 (9)

In these expressions, b is a measure of how much morethe consumer prefers product 2 over product 1; it comparesthe ‘‘wants and needs.’’ In turn, a represents how much theconsumer is aware of the superiority of the product. Thisidea comes from Hedonic theory.48–50

To illustrate the role of b, we consider a 5 1, that is, con-sumer’s perfect knowledge of both products. Then consumingk units of product 1, gives the same utility (satisfaction) asconsuming k/b units of product 2. For b 5 0.5 one needs twiceas many units to achieve the same utility level. Conversely, ifthe consumers have the same preference for each product, thatis they are indifferent, then b 5 1, and only half of the peopleknow about the new product, then consuming k units of prod-uct 1 gives the same utility by consuming ak units of product2. For a 5 0.5 one unit of product 1 give the population thesame utility as half unit of product 2. This, however, does notwork the same way for small increments around a given pointwhere both demands are different from zero, in other words,the same utility is not achieved in the simple manner describedabove. Indeed, setting du 5 0 renders:

Dd2 ¼ ab

� �q d2d1

� �1�q

Dd1 (10)

which says that the same utility is achieved by a substitutionof product 2 by product 1 in a manner that is dependent of thelevel of consumption (d1 and d2).

Under these conditions, the solution to consumer utilitymaximization is given by the following implicit equation for d1

Uðd1Þ ¼ p1d1 � ab

� �q

p2Y � p1d1

p2

� �1�q

dq1 ¼ 0 (11)

The above function has several properties. Among others:(a) it predicts d1 5 d2 when the prices are equal and when a/b 5 1, (b) it predicts a monotone decreasing value of d1with p1, which makes sense, and most important (c) it pre-dicts a monotone decreasing value of d1 with b (the larger bis the worse product 1 compares), (d) for the same prices (p15 p2), same type of products (b 5 1) one obtains

d1 ¼ aq

1�qd2, which indicates that the market is split unevenlywhen consumers are not totally aware of the new product,and (e) its elasticity of substitution is still constant. Indeed,

rES ¼ 1

1� q(12)

Although the above utility function was chosen because ofits appealing structure (elasticity of substitution), we recog-nize the need to choose the appropriate utility for each prod-uct, an exercise that should originate in marketing surveys.While there are many objections one can make to the use

above utility function, the concepts advanced will not changeif a new utility is proposed; only some expressions will. Thus,we will accept this form, to advance the point we are trying tomake about the need to incorporate the consumer behavior.

There are of course, other utility functions that lead to dif-ferent relationships between price and demand. In addition,there are other pricing methods that do not rely on the con-sumer utility maximization. For example, cost-plus pricing isa method commonly used by firms. The common thread inall forms of common pricing is that the price is determinedby the cost of the product plus an additional amount to repre-sent profit. Different forms vary in the way the surplus is cal-culated. In this option, even though the price is fixed, someprice-demand relationship is still needed.

Regardless of the type of function used, its determinationfor a new product would be challenging and often not neces-sarily very accurate. While overcoming challenges is part ofprogress I worry about overcoming uncertainties, which Ibelieve can be handled using two stage stochastic frame-works as we discuss in more detail later.

Thus, changing the product leads to changes in b and there-fore this influences sales. I discuss now a consumer preferencemodel that can be used to obtain the inferiority function b.

Consumer Preference Model

I suggested12 that the consumer preference coefficient b isgiven by the ratio the competition preference function (H2)to the new product preference function (H1):

b ¼ H2=H1 (13)

Thus, if the preference for product 2 is half of that forproduct 1, b 5 0.5. In turn the consumer preference functionis proposed to be constructed as follows:

Hi ¼Xj

xi;jyi;j (14)

In this expression the property scores yi,j of characteristicsare the contribution of property j to the preference functionof product i (like for example effectiveness, durability, feel,form, scent, and toxicity of an insect repellent, which is ourexample presented below). These scale from zero to one. Inturn, wi,j are weights, satisfying

Pjxi;j ¼ 1, which determine

the importance of each product attribute and is determinedsolely by surveys.36

These consumer related or ‘‘marketing’’ properties (con-sumer-properties as defined above) are properties that a regu-lar consumer or surveyor can relate to and do not use anyengineering jargon. The task of engineers is to connect theseproperties to physical properties or functionalities and ulti-mately to product composition or functionality or structure.This connection is the essence of our theory of productdesign. We now illustrate this briefly through an example.

Example

Consider an insect repellent to compete with an emergingcompetitor of the current market leader, a DEET-based repel-lent. It was decided that the basic active ingredient would be


Picaridin, the same as the emerging competitor. Four ingre-dients were chosen to contribute to these characteristics: pic-aridin, ethanol, aloe, and fragrance. This example was chosenfor its simplicity and because it illustrates well the relation-ships between product structure, profitability, and consumerpreferences. The optimization model is in this case is verysimple. First, manufacturing is reduced to a set of mixingtanks and we assume that the ingredients are bought. Wealso make some simplifying assumptions about the supplychain: We assume that the US divided into several regions,we consider one distribution center in each region and weassume one fixed location for the manufacturing plant.Finally, we choose only one type of consumer: middle age,camping families.

Then, we want to choose the product composition and theoptimal price to maximize the Profit11 for which we use anet present value. This has been investigated and discussedin detail by Bagajewicz.51,52 The level of demand that themodel chooses determines the associated FCI. In a simplifiedmanner, for just one product a deterministic model is as fol-lows: Let x be the composition of the product, p its price anddft the discount factor for the time period t. Then,

Maxx;p

NPV

s.t.

NPV ¼ Revenues�Manufacturing Costs

� Fixed Capital Investment

Revenues ¼Xt

dft ðDemand�pÞ

Demand ¼ Demand ðx,pÞFixed Capital Investment ¼ Fixed Capital Investment

ðDemandÞManufacturing Costs ¼ Manufacturing Costs

ðDemand, xÞTransportation Costs ¼ Transportation Costs ðDemandÞ

We omit for the moment complications like many markets,different products and process for these markets, time vary-ing parameters, the role of advertising, etc.

This scheme has been also applied to a variety of prod-ucts: Skin replacement products,53 skin lotions,54 wine mak-ing,55 novel biomedical devices.56

Consumer properties

Six important consumer properties characteristics of the re-pellent were chosen: effectiveness, durability, stickiness,form, scent, and toxicity.57 We now show how to relate thoseto concentration.

Effectiveness. A common experiment performed onrepellents is the ‘‘mosquitoes in a box’’ test. In this test, aknown mosquito population is placed inside a long rectangu-lar box. One side of the box is treated with the repellent ofinterest and at the end of a certain amount of time, the num-ber of mosquitoes on the repellent side of the box is counted.Fifty percent of the mosquito population on the repellent sidewould prove the repellent was ineffective and would corre-spond to a utility of zero. Zero mosquitoes on the repellent

side would prove the repellent was completely effective andwould correspond to a utility of 100. We now present oneresult of such experiment (Figure 4) where all data presentedcomes from informal surveys of small number of persons,performed only to advance the concept and should not to beused to make conclusions.

Figure 4a shows a graph that relates the consumer prefer-ence score of effectiveness to the common person/consumermeasurable property (% of mosquitoes). The first graphcomes from hypothetical observation (abscissa) and surveyof consumers (ordinate). What is important here is the trend.

Figure 4. Effectiveness score versus % picaridin: (a)effectiveness score versus effectiveness; (b)effectiveness versus % picaridin; (c) effecti-venss score versus % picaridin.



Figure 4b relates the measured quantity of mosquitoes to %picaridin. This second graph is also experimental in thiscase. The third graph (Figure 4c) provides the relationshipsought yeff5yeff(x).

Durability. Durability is defined as the length of timethat one dose of repellent remains effective. We assume thata great repellent, one that would have a score of 100%,would last 12 h or more, and would be best explained by alinear relationship with slope 100/12 (%/h). Next, the repel-lent durability (time) needs to be related to some physicalproperty of the repellent. This physical property is the com-position of the overall liquid mixture. For simplicity, wechose to model the situation like follows: (a) There is avapor layer of composition cis immediately close to the liq-uid that is in equilibrium with the liquid composition, that is,cis 5 pis / RT 5 xiP

0i (T)ci / PRT, (b) the rate of removal of

the mixture from the layer is assumed to be given by a natu-ral convection mass transfer coefficient (although a moreelaborate diffusion model can be constructed), that is:N ¼ h

Picis, and (c) replenishment of the vapor phase to

reach equilibrium is considered instantaneous. Therefore, onecan write dmi

dt ¼ �ANi ¼ �AcisN, and M ¼ Pimi, where A is

the area ; so after substitution one obtains a differential equa-tion for the mass of each component in the liquid as a func-tion of composition, which can be integrated numericallyusing mi 5 m0

i at t 5 0. For the mass transfer coefficient wehave used a correlation for forced convection turbulent masstransfer on a flat plate (k�q ¼ 0:0365N0:8

Re;LDL). We understand

this model can be enhanced substantially, but we chose tokeep it simple and only for the purpose of being able toadvance the conceptual approach we are presenting. Results,which illustrate the concept, are shown in Figure 5 includingthe final durability score.

Stickiness. The first step in relating utility to stickinesswas assigning qualitative descriptions to levels of stickinesspreference (Figure 6a), which comes directly from consumersurveys. Then, we relate these levels of stickiness to somemeasurable physical property through a ‘‘Paper Test’’ (Figure6b). To perform this test, a person applies repellent of a spe-cific formulation to the underside of his arm and places atwo-inch-by-two-inch piece of paper on the applied area. Thethickest piece of paper that sticks to the applied area anddoes not fall off determines the stickiness of the repellent.Thickness of paper, or basis weight, is measured by theweight of 500 sheets of that type of paper. For example, afull sheet of 50-pound paper would weigh 1/500 of 50pounds, or one tenth of a pound. The next step is to relatethis consumer test to a physical property of the repellent for-mula. Ethanol and picaridin are nonsticky, so only aloe canbe related to the feel consumer test. For simplicity, weassumed each contributes independently of the other. Allresults are shown in Figure 6.

Scent. To construct the scent utility function, the con-sumer determines how satisfying each fragrance scent strengthwould be to them. In addition, alcohol also contributes toscent but negatively. Thus, for we compute the total scentscore using the weighted average of two preference scores:

Yi;scent ¼ Yi;ethanolxi;ethanol þ Yi;fragrancexi;fragrancexi;ethanol þ xi;fragrance

(15)

Figure 7 shows the results: The first two figures show thescores for fragrance and ethanol as a function of consumerperception. The most preferred point is where the repellenthas only a trace scent, and it decreases for any change instrength. Ethanol, in turn, has an increasing negative effectfor the whole range. A linear relationship between concentra-tion and scent power is used for both species (100% corre-sponding to overpowering and 0% to none).

Form. There are two forms of repellent available to con-sumers-lotion or spray. The most important physical property

Figure 5. Durability score versus % picaridin: (a) dura-bility score versus repellent durability; (b) du-rability versus % picaridin; (c) durability scoreversus % picaridin.



to ascertain this form is the mixture viscosity (surface tensionwould address droplet size) because it determines if the prod-uct will be flowing free enough to be a spray. If it is toothick, it will be a gel or lotion. Liquids with a kinematic vis-cosity over 75 centistokes will be too thick to be sprayed bya finger pump, a typical packaging for insect repellent. Thevalues for dynamic viscosity are known or estimated foreach of the materials. For any mixture, the resulting dynamicmixture viscosity is calculated with the Grunberg and Nissanmethod58 and converted to kinematic viscosity.

The form score is derived from consumer preferences. Forexample, if z% of consumers prefer spray repellent over thelotion form, a repellent in spray form would have ‘‘100%score’’ to z% of consumers, but smaller, 50% in our case, tothe other (1 2 z)%. Thus, a spray repellent would have anoverall consumer preference score of zcs 5 z 1 0.5*(1 2 z).Conversely, a repellent in lotion form would have a score ofzcl 5 (1 2 z)1 0.5z. Finally, the relationship between viscos-ity and utility can be expressed with an ‘‘If . . . then. . ..’’ state-ment giving the utility for any mixture viscosity, i.e. ‘‘If kine-matic viscosity is less than 75 centistokes, utility is zcs %; ifkinematic viscosity is more than 75 centistokes, utility is zcl.’’

Toxicity. The major benefit of a picaridin-based repellentis the decreased health risk compared to DEET-based repel-lents. A consumer preference score should be based on thedanger to health that is associated with each component. Asthe risk increases, consumer happiness will decrease; this ismodeled as a linear relationship. The risk associated with eachcomponent is derived from the National Fire Protection Asso-ciation (NFPA) Health Hazard rating, often found on MaterialSafety Data Sheets (MSDS). The NFPA ratings are as followsfor each material: DEET-2; picaridin-1; ethanol-1. Results areshown in Figure 8. A linear relationship is used to describethe NFPA toxicity score as the concentration of each.

Weights. These are given in Table 2 and were again ob-tained using an informal survey of a small number of people.

The best product

When consumer preference (H1) was maximized, which isequivalent to seeking the minimum of b (because H2 isfixed), the result suggested a product that is 98.21% picari-din, 1.79% ethanol, 0% aloe, and 0% fragrance with a bvalue of 0.524. This makes sense because of the weights

Figure 6. Stickiness score versus aloe concentration: (a) stickiness score versus stickiness perception level; (b)paper basis weight versus stickiness perception level; (c) stickiness (paper weight) versus % aloe; (d)stickiness preference score versus % aloe.

[Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]


used. Such a product is not profitable as it will be shown inthe next section.

The most profitable product

The competitor sales price is $90 and has a formulation of7% picaridin, 30% ethanol, and the rest water. We used avalue of q 5 0.76 and a value of a 5 0.9, that is, almost per-fect knowledge of the new product. Figure 9 shows NPVcurves as a function of the proposed price for the new product.We see that the optimal value of beta for this market with theaforementioned weights is b 5 0.67, which gives an NPV ofalmost 12,000,000. This value of b corresponds to a concen-tration of 40% picaridin, 58% ethanol, 1% aloe, and 1% fra-grance (these last two having reached their imposed lowerlimit). Interestingly, the curve for b 5 0.76 shows a peak at$90 (same price as the competition) with an NPV of around10,700,000 (the corresponding concentration is 30% picaridin,63% ethanol, 3.3% aloe, and 3.3% fragrance. Lower values ofb (0.61 and 0.59) show remarkable lower profit. For values ofb lower than 0.59, the NPV is smaller and does not achieve amaximum in the range of prices chosen, showing increasingmonotonicity and crossing from negative to positive NPV’s at

larger prices. We consider large prices unrealistic and weexpect the above consumer model to break down when pricesare so different. This will be object of future work. Larger val-ues of b result also in a lower profit anyway.

Multiple Markets

We also investigated the effect of dealing with differentmarkets. We assumed the plant located in Little Rock, ARand the markets corresponding to Table 3. The correspondingcomposition is 23.8% picaridin, 0% aloe, 65% alcohol, and11.2% fragrance. We programmed the model in excel andused solver, so the results may not be globally optimal, butare very useful to illustrate the intricacies of the results. Theoptimal price is $90, that is, the same as the competition(Figure 10). This is not surprising because most of the valuesof b are close to one (Table 4). When ROI was maximized,the result is different (Figure 11). The composition is ratherdifferent, 26% picaridin, 0% aloe, 65% alcohol, and 9% fra-grance and the suggested price is $108. The values of b aresimilar to the previous case, but the capital investment ismuch smaller (around $1 million vs. $1.35 million for maxi-mum NPV). Finally, the discontinuities in all graphs stem

Figure 7. Scent preference score from fragrance and ethanol: (a) fragrance preference versus scent power percep-tion; (b) ethanol scent preference score versus scent power; (c) scent preference versus % fragrance; (d)scent preference versus % ethanol.



from the fact that when a market had revenues smaller thantransportation costs, the market was eliminated. For example,at $108, Eugene is responsible for the discontinuity.

Unresolved/Unexplored Issues

We now explore some issues that have been left aside inthe above example but are important to discuss. They will beaddressed in future work.

Awareness function

The function a is usually a function of time. It can beassumed to be of sigmoidal form, with the public starting ata 5 0 or a slightly larger value if there is some element thatmakes the public instantly aware, like being suddenly presentin some store shelves, passing through a growth periodreaching a saturation level of some sort. The curve isaffected by advertising, which in turn, has a cost. Figure 12shows two such curves.

Advertising

To effectively advertise a product, service, or good, thefollowing concerns must be addressed and determined. Theseconcerns include: (a) the size of the total advertising budget,

Figure 8. Toxicity score versus ethanol and picaridin concentration: (a) toxicity score versus Toxicity; (b) NFPA tox-icity versus NFPA toxicity description; (c) NFPA toxicity description versus % ethanol 1 picaridin; (d) con-sumer preference toxicity score versus % thanols 1 picaridin.


Figure 9. Profit as a function of quality and price, onemarket.


Table 2. Weights of the Preference Function

Characteristic Weight

Effectiveness 0.29Durability 0.24Feel 0.19Form 0.14Toxicity 0.09Scent 0.05


(b) the allocation of this budget to marketing areas, (c) theallocation of the individual market area budgets amongmedia, (d) the timing of advertising, (e) the theme of thecampaign and, (f) the effort to be invested in a campaign.Figure 13 depicts the basic advertising trend for a genericproduct, service, or good. During the beginning of advertis-ing, there is a linear trend between the advertising rate andsales of the product. Once the product begins to gain popu-larity, the sales reach a threshold, and the trend betweensales and advertising rate is no longer linear. Soon, the prod-uct will begin to saturate the market and the product reachesits height in popularity. At this point, with an increasedadvertising rate, the product begins to oversaturate the mar-ket. As a result, sales begin to decline (see Rao59).

The relationship between sales and advertising efforts canalso be modeled through the influence of advertising on theawareness and superiority functions (a and b). It might alsoaffect the consumer’s budget (Y). Finally, the level of adver-tising needed should be suggested by the model becauseadvertising increases the aforementioned parameters, but alsoincreases marketing costs.

Superiority function

The function b is also subject to change in time and beaffected by advertisement. We envision a natural decaythrough time, which sometimes has to do with technologicaladvances, and which advertising can partially prevent.

Time horizon

With the model parameters (a, b, q, Y) naturally changingwith time the right model should include the time horizon.

Pricing

Pricing of certain products is subject to several types ofdiscount structures that need to be incorporated. In addition,one should think of incorporating in the model the possibilityof picking different prices for different times.

Profit function

As we have explored measuring profit through NPV andROI, there are other measures or restrictions one might wantto use like the cash position of the company, the share holdervalue strategy, liquidity ratios, etc. A recent paper by Prof.

Table 3. Weight Factors for Many Markets

Market

Weighted Averages

Effectiveness Durability Feel Form Toxicity Scent

Pittsburg 0.286 0.238 0.190 0.143 0.095 0.048Phoenix 0.238 0.286 0.190 0.143 0.095 0.048Eugene 0.190 0.238 0.286 0.143 0.095 0.048Denver 0.143 0.190 0.238 0.286 0.095 0.048Salt Lake City 0.143 0.095 0.048 0.286 0.238 0.190Lubbock 0.095 0.143 0.190 0.238 0.286 0.048Kansas City 0.048 0.095 0.143 0.190 0.238 0.286Sacramento 0.095 0.143 0.238 0.190 0.286 0.048Indianapolis 0.190 0.143 0.095 0.048 0.286 0.238Jacksonville 0.238 0.190 0.143 0.095 0.048 0.286Albany 0.143 0.190 0.286 0.238 0.095 0.048Billings 0.095 0.048 0.286 0.238 0.190 0.143Baton Rouge 0.048 0.286 0.238 0.190 0.143 0.095St Paul 0.238 0.190 0.048 0.095 0.143 0.286Memphis 0.143 0.190 0.048 0.095 0.286 0.238Charlotte 0.048 0.095 0.286 0.238 0.143 0.190

Figure 10. Profit as a function of quality and price,many markets.


Table 4. Values of b for Different Markets

Market b (Best ROI) b (Best NPV)

Pittsburg 0.798 0.829Phoenix 0.784 0.816Eugene 0.855 0.885Denver 0.901 0.924Salt Lake City 1.014 1.022Lubbock 0.974 0.990Kansas City 1.091 1.095Sacramento 0.979 0.997Indianapolis 0.999 1.014Jacksonville 0.907 0.931Albany 0.906 0.931Billings 1.059 1.071Baton Rouge 0.890 0.915St. Paul 0.914 0.934Memphis 0.970 0.985Charlotte 1.049 1.061


Umeda discusses some of the connections between projectdecisions and corporate strategies.60

Supply chain issues

The examples above do not include illustrations about theintricacies and effect on the product design as well as thechoice of markets that are associated to the design and costsof the supply chain. Models such as the one proposed byTsiakis et al.61 need to be used. Lavaja et al.32,62 exploredsome of these issues and showed that in the context of multi-ple markets and multiple prices a new commodity can giverise to more than one manufacturing location. Inventory lev-els are also usually ignored at the planning stages. Theyplay, however, an important role in managing financialrisk.63

Budgeting

Cash management is of paramount importance to deter-mine the real profitability of a project. In most chemical en-gineering models cash flow is calculated assuming one set ofconditions, which are kept the same for the life of the pro-ject. Even textbooks of recent update, like Peters et al.64 con-tinue to use this type of assumptions. Recently some workhas started to depart from this formulation, proposing to fol-low the cash flow throughout time in more detail.32,62,65–67

Adding cash flow balance constraints allow to determinewhen and what portion of proceeds is returned as pure profitto the investors or to the company’s general account, whatportion is used to finance expansions (if an when this ismore advantageous than fresh capital investment). One canconsider the use of short term market instruments such astreasury bonds, etc, as it is suggested in cash managementmodels.68 Finally, capital investment limits may change thesolution.

Contracts

Contracts play an important role in defining the shape ofthe enterprise. Regular or option contracts help determineand manage the financial risks.47,63,69

Competition

The model outlined above considers a static competitor,which does not react to the new market. In reality, one needsto assume a dynamic condition and has to include the com-petitor model, that is, a model that will determine the com-petitor’s price (or competitors’ prices). Although an extensivereview can be made of several models (perfect competition,monopoly, oligopoly, etc), we leave this matter unexploredin this article.

Uncertainties and financial risk

Many parameters of the model are uncertain, truly uncer-tain, like a, b, q, Y. Moreover, the uncertainties behind b canbe related to uncertainties in the weights that define the con-sumer preference function as well as the consumer perceptionof each of the function scores ( y). There are, in general two al-ternative routes, one is that of building full blown two (ormulti)-stage stochastic programming models for expectedprofit maximization70,71 and financial risk,47 the dynamic pro-gramming approach.72 We prefer the former approach. In addi-tion, to deal with financial risk, we proposed to use the frame-work advanced by Barbaro and Bagajewicz47 and the riskaverse curve selection proposed by Aseeri and Bagajewicz.69

Figure 11. ROI as a function of quality and price, manymarkets.


Figure 12. Awareness functions.


Figure 13. Sales as a function of advertising.


Computational issues

As stated earlier, many products may be associated withcomplex industry structures (like pharmaceutical drugs) whichwould make the models too complex to solve. We envisionthat these models could be complex MINLP’s not more com-plex than the ones we are currently solving is several synthesisproblems with multiperiod structure. We point out, however,that at least in the case of the design of one product asopposed to multiple products, a nice decomposition scheme ispossible. In fact it has been applied in this article. Indeed, onecan actually enumerate the product consumer preference(through the parameter b), by selecting different discrete val-ues, which leads to a nice decomposition: With b fixed, onecan solve for the product structure by minimizing its manufac-turing cost. This would be equivalent to applying the currentproposed methodology of meeting consumer ‘‘needs.’’ Withthe costs defined, one can assess the NPW easily without theneed for any optimization, unless the supply chain is to bedesigned, but this problem is also known as solvable to a greatextent. When multiple products are designed, one needs torepeat this scheme for values of the parameter b correspond-ing to the different products. This is still doable.

Conclusions

The basic message that this article conveys is that pricingand microeconomics, as well as with supply chain, processsynthesis and finances are needed when one wants to designnew products. A framework was developed in which all theelements of new product commercialization, namely, theproduct composition/structure/functionality, the manufactur-ing investment and costs, the associated supply chain and theconsumer behavior with respect to price product and priceare taken into account in a model that determines all the pa-rameters of the subsystems involved, form the product struc-ture to the choice of markets and the price of the product ineach market.

The timing to construct such a model after the first ideasof a new product are laid out is also arguable. Some wouldargue that it is not practical to do so too early in the design,especially because of the effort (and associated cost)involved. Because one cannot only target consumer ‘‘needs’’and how to meet them, but also needs to worry about profit-ability, I argue that the analysis suggested in this articleneeds to be conducted, whether quantitatively as we propose,or in some other heuristic form.

Acknowledgment

Kamila Sobeslavova also helped in the preparation of this manuscript.

Literature Cited

1. Westerberg AW, Subrahmanian E. Product design. Comput ChemEng. 2000;24:959–966.

2. Cussler EL, Moggridge GD. Chemical Product Design, 1st ed. Cam-bridge University Press, 2001.

3. Hill M. Product and process design for structured products. AIChEJ. 2004;8:1656–1661.

4. Seider WD, Seader JD, Lewin DR. Product and Process DesignPrinciples. New York: Wiley, 2004.

5. Stephanopoulos G. Invention and innovation in a product-centeredchemical industry: general trends and a case study. In AIChE Con-ference, 55th Institute Lecture, San Francisco, November 2003.

6. Cussler EL. Chemical product design and engineering (PlenaryTalk). In AIChE Annual Conference, Paper 430a, San Francisco,November 2003.

7. Cussler EL, Moggridge GD. Chemical product design. In FOCAPDConference, Princeton, USA, July 2004.

8. Costa R, Moggridge GD, Saraiva P. Chemical product engineering:a future paradigm. Chemical Engineering Process. April 10–13,2006.

9. Costa R, Moggridge GD, Saraiva P. Chemical product engineering:an emerging paradigm within chemical engineering. AIChE J.2006;52:1976–1986.

10. Ng K, Gani R, Dahm-Johansen K. Chemical Product Design:Towards a Perspective Through Case Studies. Computer AidedChemical Engineering Series 23. Amsterdam: Elsevier, 2006.

11. Bagajewicz M. Integration of process systems engineering and busi-ness decision making tools: financial risk management and otheremerging procedures. In: Galan M, del Valle EM, editors. ChemicalEngineering Trends and Developments. pp. 323–377. Chichester,U.K.: Wiley, 2005.

12. Bagajewicz MJ. On the integration of process and product designwith microeconomics and finances. In Enpromer 2005, 2nd MercosurCongress on Chemical Engineering, 4th Mercosur Congress on Pro-cess Systems Engineering, Rio de Janeiro, Brazil, August 2005.

13. Wibowo C, Ng KM. Product-oriented process synthesis and develop-ment: Creams and pastes. AIChE J. 2001;47:2746–2767.

14. Joback K, Stephanopoulos G. Searching spaces of discrete solutions:the design of molecules possessing desired physical properties. AdvChem Eng. 1995;21:257–311.

15. Sahinidis NV, Tawarlamani M. Applications of global optimizationto process molecular design. Comput Chem Eng. 2000;24:2157–2169.

16. Siddhaye S, Camarda KV, Topp E, Southard MZ. Design of novelpharmaceutical products via combinatorial optimization. ComputChem Eng. 2000;24:701–704.

17. Siddhaye S, Camarda KV, Southard MZ, Topp E. Pharmaceuticalproduct design using combinatorial optimization. Comput ChemEng. 2004;28:425–434.

18. Kier LB, Hall LH. Molecular Connectivity in Chemistry and DrugResearch. New York: Academic Press, 1986.

19. Achenie LEK, Sinha M. Interval global optimization in solventdesign. Reliable Comput. 2003;9:317–338.

20. Douglas JM. Conceptual Design of Chemical Processes. New York:McGraw Hill, 1988.

21. Smith R, Linnhoff B. The design of separators in the context ofoverall processes. Trans IChemE ChERD. 1988;66:195.

22. Smith R. Chemical Process Design. New York: McGraw Hill, 1995.23. Biegler LT, Grossmann IE, Westerberg AW. Systematic Methods of

Chemical Process Design. New Jersey: Prentice Hall, 1997.24. Camarda KV, Maranas CD. Optimization in polymer design using

connectivity indices. Ind Eng Chem Res. 1999;38:1884–1892.25. Ostrovsky G, Sinha M, Achenie LEK. On the solution of mixed-in-

teger nonlinear programming models for computer aided moleculardesign. Comput Chem. 2002;26:645–660.

26. Evans L. Achieving operational excellence with information technol-ogy. In AICHE Spring Meeting, New Orleans, March 2003.

27. Gani R. Chemical product design: challenges and opportunities.Comput Chem Eng. 2004;28:2441–2457.

28. Cordiner JL. Challenges for the PSE community in formulations.Comput Chem Eng. 2004;29:83–92.

29. Grossmann IE, Westerberg AW. Research challenges in process sys-tems engineering. AIChE J. 2000;46:1700–1703.

30. Grossmann IE. Challenges in the new millennium: product discoveryand design, enterprise and supply chain optimization, global lifecycle assessment. In PSE 2003, 8th International Symposium on Pro-cess Systems Engineering, China, 2003.

31. Breslow R, Tirrell MV, (co-chairs) Barton JK, Barteau MA, BertozziCR, Brown RA, Gast, AP, Grossmann IE, Meyer JM, Murray RW,Reider PJ, Roush WR, Shuler ML, Siirola JJ, Whitesides GM, Wol-nyes PG, Zare RN. Beyond the molecular frontier. challenges forchemistry and chemical engineering. In Committee on challenges forthe chemical sciences in the 21st century, The United StatesNational Academies Press, 2003.


32. Lavaja J, Adler A, Jones J, Pham T, Smart K, Splinter D, Steele M,Bagajewicz M. Financial risk management for new products: consid-erations of plant location, pricing and budgeting. In Proceedings ofthe AIChE Annual Meeting, Austin, Texas, 2004.

33. Kahn KB. PDMA Handbook of New Product Development, 2nd ed.New York: Wiley, 2004.

34. Ng K. MOPSD: a framework linking business decision-making toproduct and process design. Comput Chem Eng. 2004;29:51–56.

35. Achenie L, Venkatasubramanian V, Gani R. Computer Aided Molec-ular Design: Theory and Practice. Amsterdam: Elsevier Science,2002.

36. Moskowitz HR. Consumer Testing and Evaluation of Personal CareProducts. New York: Marcel Dekker, 1996.

37. Trout J. ‘‘Positioning’’ is a game people play in today’s me-too mar-ket place. Ind Market. 1969;54:51–55.

38. Ries A, Trout J. Positioning, The Battle for your Mind. New York:Warner Books, McGraw-Hill, 1981.

39. Trout J, Rivkin S. The New Positioning: The Latest on the World #1Business Strategy. New York: McGraw Hill, 1996.

40. Marn MV, Roegner EV, Zawada CC. The Price Advantage. NJ:Wiley, 2004.

41. Guillen G, Bagajewicz M, Sequeira SE, Espuna A, Puigjaner L.Management of pricing policies and financial risk as a key elementfor short term scheduling optimization. Ind Eng Chem Res.2005;44:557–575.

42. Dorward N. The Price Decision: Economic Theory and BusinessPractice. London: Harper & Row, 1987.

43. Varian H. Microeconomic Analysis. New York: W.W. Norton &Company, 1992.

44. Mas-Collel A, Whinston M, Green J. Microeconomics Theory.Oxford: University Press, 1995.

45. Hirshleifer J, Hirshleifer D. Price Theory and Applications. New Jer-sey: Prentice Hall, 1998.

46. Pekny J. Algorithm architectures to support large-scale process sys-tems engineering applications involving combinatorics, uncertainty,and risk management. Comput Chem Eng. 2002;26:239–267.

47. Barbaro AF, Bagajewicz M. Managing financial risk in planningunder uncertainty. AIChE J. 2004;50:963–989.

48. Rosen S. Hedonic prices and implicit markets: product differentia-tion in pure competition. J Polit Econ. 1974;82:34–55.

49. Epple D. Hedonic prices and implicit markets: estimating demandand supply functions for differentiated products. J Polit Econ.1987;95:59–80.

50. Kahn S, Lang K. Efficient estimation of structural hedonic systems.Int Econ Rev. 1988;29:157–166.

51. Bagajewicz M. Effect of pricing, advertisement and competition inmulti-site capacity planning. In Proceedings of the ESCAPE 2005,Barcelona, Spain, May 2005.

52. Bagajewicz M. Integration of process systems engineering and busi-ness decision making tools: financial risk management and otheremerging procedures. In: Galan M, del Valle EM, editors. ChemicalEngineering Trends and Developments. pp. 323–377. Chichester,U.K.: Wiley, 2005.

53. Tang H, Azzarello J, Ashaye T, Fairbanks B, Hargis M, Williams P,Shaw B, Klink I, Sikavitsas V, Bagajewicz M. Modeling of theFDA approval process: connections between financial risk, early de-

cision making and future pricing. In Proceedings of the AIChE An-nual Meeting, Austin, Texas, 2004.

54. Bagajewicz M, Lopez H, Sposato E. Use of pricing and consumersatisfaction measures in consumer product design. In AIChE AnnualMeeting, Cincinnati, November 2005.

55. Bagajewicz M, Kerr S, Frow M. Engineering wine. In AIChE An-nual Meeting, San Francisco, November 2006.

56. Bagajewicz M, Froude V, Burdett E, Shreve M, Clemente-Harl E,Martin M, Sikavitsas V. Risk management in the development ofnovel biomedical devices and vaccines. In AIChE Annual Meeting,San Francisco, November 2006.

57. Ashley E, Doman S. Insect Repellent Design: Researching Alterna-tives to DEET. University of Oklahoma Chemical EngineeringAdvance Design Class Project (Available upon request to theauthor), 2006.

58. Reid RC, Prausnitz JM, Sherwood TK. The Properties of Gases andLiquids, 4th ed. New York: McGraw Hill, 1987.

59. Rao AG. Quantitative Theories in Advertisement. New York: Wiley,1970.

60. Umeda T. A conceptual framework for the process system synthesisand design congruent with corporate strategy. Ind Eng Chem Res.2004;43:3827–3837.

61. Tsiakis P, Shah N, Pantelides CC. Design of multi-echelon supplychain networks under demand uncertainty. Ind Eng Chem Res.2001;40:3585–3604.

62. Lavaja J, Adler A, Jones J, Pham T, Smart K, Splinter D, Steele M,Bagajewicz M. Financial risk management for investment planningof new commodities considering plant location and budgeting. IndEng Chem Res. 2006;45:7582–7591.

63. Barbaro AF, Bagajewicz M. Use of inventory and option contractsto hedge financial risk in planning under uncertainty. AIChE J.2004;50:990–998.

64. Peters MS, Timmerhaus KD, West RE. Plant Design and Economicsfor Chemical Engineers, 5th ed. New York: McGraw Hill, 2003.

65. Romero J, Badell M, Bagajewicz M, Puigjaner L. Integrating budg-eting models into scheduling and planning models for the chemicalbatch industry. Ind Eng Chem Res. 2003;42:6125–6134.

66. Badell M, Romero J, Huertas R, Puigjaner L. Planning, schedulingand budgeting value-added chains. Comput Chem Eng. 2004;28:45–61.

67. Powell S, Hodge S, Spencer N, Godwin J, Bagajewicz M. Simulta-neous modeling of location, advertisement and competition in invest-ment/capacity planning with risk management. In Proceedings of theAIChE Annual Meeting, Austin, Texas, 2004.

68. Orgler YE. Cash Management. California: Wadsworth, 1970.69. Aseeri A, Bagajewicz M. New measures and procedures to manage

financial risk with applications to the planning of gas commercializa-tion in Asia. Comput Chem Eng. 2004;28:2791–2821.

70. Eppen GD, Martin RK, Schrage L. A scenario approach to capacityplanning. Oper Res. 1989;37:517–527.

71. Ahmed S, Sahinidis NV. Robust process planning under uncertainty.Ind Eng Chem Res. 1998;37:1883–1892.

72. Cheng L, Subrahmanian E, Westerberg AW. Design and planningunder uncertainty: issues on problem formulation and solution. Com-put Chem Eng. 2003;27:781–801.

Manuscript received Dec. 28, 2006, and revision received Aug. 6, 2007.


on the role of microeconomics, planning, and finances in ... the role... · on the role of...

Documents